MIT study finds OpenAI models solve about 35% of unpublished theory problems
Overview
A study led by Neil Thompson at MIT finds OpenAI models solved about 35% of unpublished theory problems tested from December 2025 to May 2026, up from about 2% between mid-2024 and mid-2026.
The problems covered probability, statistics theory, economic theory, high-energy theory, gravitation and cosmology, and quantum physics, according to Wired.
The researchers caution that AI may make mistakes and that experts are still needed to verify the solutions.
Written by AI from the articles below · updated Oct 14, 2:47 PM ET
Check the sources:
Article timeline
The articles in this story. Times are ET.
- Wired · AINewsNeil Thompson study finds AI models now solve many theory problems in other sciences
AIA study led by Neil Thompson at MIT tested OpenAI models on unpublished theory problems from December 2025 to May 2026. Across probability, statistics theory, economic theory, high-energy theory, gravitation and cosmology, and quantum physics, the share solved rose from about 2 percent between mid-2024 and mid-2026 to about 35 percent. The researchers caution that AI may make mistakes and that experts are still needed to verify solutions.
Heat trend
- Current comparable heat
- 9
- Peak comparable heat
- 10Oct 14, 3:00 PM ET
- Change in past 24 hours
- –
The trend compares only continuously observed participants, so its scope may be smaller than the current heat total. Hover or tap for hourly heat; use arrow keys with a keyboard.